The brain is a common site of treatment failure for many primary tumors. the ability to achieve local control of brain metastases is critical, as patients whose lesions respond to radiation therapy live twice as long as those who do not. Standard treatment with whole brain radiation therapy is of limited efficacy and recent attention has focussed on techniques such as radiosurgery which deliver a relatively high dose to the region of the tumor. The initial results obtained with radiosurgery have been extremely promising. To more accurately characterize the treatment response and understand the mechanisms of failure, requires the development of clinical examinations which can provide reliable quantitative assessments of changes in tumor volume and metabolism. In this study, we will apply volume Magnetic Resonance Imaging and spectroscopy (MRI/MRS) techniques, in conjunction with advanced data analysis algorithms to quantify changes in tissue morphology and function in serial examinations. This novel approach to characterizing treatment response will be applied to eighty patients with brain metastases from lung carcinoma, colon carcinoma, breast carcinoma or melanoma. The treatments considered will be whole brain radiation therapy, radiosurgery alone or combined whole brain radiation therapy and radiosurgery. Examinations will be performed pre-treatment, at the end of treatment, one month after therapy and at three monthly follow-ups. The MRI data acquired will comprise volume or thin multi-slice images with T1-weighting, proton density weighting, T2-weighting and gadolinium enhanced T1-weighting. These images will be reformatted, aligned and segmented into components with intensities characteristic of edema, contrast enhancing tumor, grey matter, white matter and csf. Volumes of edema and contrast enhancing tissue will be estimated and used to assess changes in morphology occurring in response to therapy. The MRS data will comprise water suppressed 3-D localized 1H spectra acquired from a PRESS selected region of 100-300cc. the aim will be to obtain spectra from 1-3cc voxels, including both regions of normal brain tissue and pathology, in addition to producing images of the spatial distribution of N-acetyl aspartate, creatine, choline and lactate. Preliminary data suggest that the variations in levels of these metabolites can be used to distinguish normal brain parenchyma from recurrent tumor and radiation necrosis. Direct comparisons between the spectral data and segmented images will allow us to test hypotheses concerning the relationship between changes in tissue morphology and function in response to therapy. In addition to the MRI/MRS data, we will obtain FDG positron emission tomography examinations from 30 of the 80 patients. Comparison of the PET and MR data will indicate whether the two modalities are complementary in assessing treatment response. Radiosurgery patients will also have treatment planning MR and CT studies and all patients will have three monthly followup clinical MR or CT. The serial MRI/MRS, the PET and clinical MR/CT data will be aligned either manually or by automated surface matching. Direct comparisons will be made between the information obtained from each modality in order to establish the clinical significance of the MRI/MRS examination and data processing. In the long term, we anticipate that it will be possible to classify tumors based upon their functional and morphological characteristics, predict which radiosurgery dose levels are effective and determine which tumors require immediate additional therapeutic intervention.